Last updated October 25, 2026 · Edition I

Working With Claude: The Place to Stand Field Guide

Refreshed every two weeks from Anthropic's own release notes, the Anthropic Academy, and the few tutorials that earn a recommendation. Written for owners, partners, and staff who want results, not a computer science lesson.

I. Start Here: What Changed This Month

Three things matter if you have not opened Claude in a few weeks.

1. Cowork is now just Claude. On September 16, 2026 Anthropic merged Cowork into the main Claude app. There is no longer a mode to pick. You type what you need in one box and Claude decides whether that is a quick answer or a multi-step job with files. Everything you had in Cowork (projects, skills, connectors, scheduled tasks, artifacts, memory) carried over. Depending on your plan you may still see a Cowork option for a while: Pro and Max got the new experience first, Enterprise on October 23, Team is in beta. Once an account moves it does not switch back.

2. Your work runs in the cloud and keeps going when your laptop is closed. Since October 6, new tasks on Pro and Max run on Anthropic's servers, so the files those tasks work on are processed there too. You can start a job at your desk and check it from your phone. Scheduled tasks no longer need a computer left on. Two things still need the desktop app open on your machine: reading folders on your computer, and anything that uses your browser or your screen. Enterprise owners turn cloud sessions on themselves; they are off by default there.

3. Claude now makes real documents, decks and designs in any conversation. Claude Docs, Claude Slides and Claude Design are built into every chat. You get an editable document, a deck with speaker notes, or a visual on a canvas, and you can export to Word, PowerPoint, PDF or Google formats. The one setup step that makes these look like yours rather than generic: build a design system first (your colors, fonts and logo) and point Claude at it.

Also new: four model releases since September (Fable 5.1, then Opus 5.5, Sonnet 5.5 and Haiku 5.5), and Claude Duet, a separate proactive assistant in a waitlisted research preview for US Pro and Max accounts. Details in section IV.

II. How Claude Works Now, On One Page

You do not need to understand the machinery. You need to know the seven parts and what each one is for.

Part	What it is, in plain words	What it is for
A conversation	The one box where you type. Claude decides whether to answer or to go and do the work.	Everything starts here.
Projects	A folder of files, standing instructions and memory for one ongoing piece of work.	Keeps a client, a department or a recurring job in one place so you never re-explain it.
Connectors	Permissions that let Claude read (and sometimes write to) your email, calendar, files, CRM, accounting or Slack.	So Claude works from your real information instead of what you paste in.
Skills	A saved recipe for a job you do repeatedly, written once and reused.	Consistency. Your proposal, report or triage in your format and voice, run after run. You still read the result.
Plugins	A bundle of skills, connectors and helpers for a whole function (sales, finance, small business, legal).	Install one and get a working toolkit instead of building skills one at a time.
Scheduled tasks	A skill or instruction that runs on its own: hourly, daily, weekly or on demand.	The Monday briefing, the Friday report, the invoice chase, done before you arrive.
Memory	What Claude keeps about you between conversations: editable topics you can review and delete.	Less repetition. On by default for individuals, off by default for Team and Enterprise until an owner turns it on.

Two settings worth knowing. Permissions on each conversation are Auto (Claude checks each action itself; the default on Pro, Max and Team) or Manual (you approve each step; the default on Enterprise). Instructions for Claude in Settings hold your standing rules: tone, formats, what to always or never do.

III. The Seven Habits of People Who Get Results

These come from Anthropic's own guidance and from the practitioners whose tutorials I recommend below. They agree on more than you would expect.

I. Teach it your business before anything else. Every good setup starts with context: who you are, how you write, who your people are, what you sell, what you never say. Anthropic's small-business plugin does this with an onboarding step that saves a business profile every later skill reads. Do this first.

II. Do the job once with Claude, then turn it into a skill. People describe their processes badly and perform them well. So do the task together, correct it until it is right, then press "turn into skill" or use "record a skill" (Claude watches your screen while you do it and writes the recipe). Do not write skills in the abstract.

III. Keep skills small and corrective. One job per skill. The best skills hold only what Claude gets wrong on its own: your exceptions, your format, your three good examples. Treat them as living documents and refine after real use. Bigger is not better; more context makes Claude worse, not smarter.

IV. Give a job only what it needs. Connect the tools a task actually uses and nothing more. Every connector and skill left switched on costs you usage on every run and widens what Claude can reach. Check your usage (type "explain usage" or the slash-usage command) and switch off what you do not use.

V. Draft, never send. The rule that keeps you safe: Claude prepares, you approve. Anything that sends, posts, pays or deletes should need your yes. Put that rule in a setting, not just in a polite instruction. Start every scheduled task in draft mode and promote it only after you have read a few runs.

VI. Schedule what recurs, then read the output. The payoff of all this setup is the Monday brief, the weekly report and the follow-up sweep running without you. Review the results in the Scheduled area, pause what is not earning its keep, and tighten the instructions over time.

VII. Describe the destination, not the route. When you ask for something, say who it is for, what result matters, what format and how long. Hand over the whole task and let Claude propose the steps. Ask it to restate the request and list its questions before it starts. Then judge the result the way you would judge a new hire's first draft.

IV. What Anthropic Shipped Lately

The dated list, newest first, in plain English. Links in the Sources section.

Date	What shipped	Why you care
Oct 23, 2026	Unified Claude generally available on Enterprise; Efficient mode; spend limits by group and member	Admins can roll it out, cap cost, and keep routine work on cheaper models.
Oct 22, 2026	Agent Skills donated to the Linux Foundation's Agentic AI Foundation	Skills are becoming an open standard supported by 70+ platforms, so yours should work outside Claude. Test that before you rely on it.
Oct 21, 2026	Claude Duet (research preview)	A separate app where Claude watches your connected accounts, surfaces what needs attention and checks before acting. Waitlist, US Pro and Max only, no Team or Enterprise date yet.
Oct 15, 2026	Claude Docs on by default for Enterprise	Living, shareable documents in any conversation.
Oct 7, 2026	Claude Haiku 5.5; monthly API credits for Max and Team	The cheap, fast model for high-volume routine work.
Oct 6, 2026	Cloud-first sessions on Pro and Max	Work keeps running with the laptop closed; see section I.
Sep 28, 2026	Claude Sonnet 5.5	Faster, cheaper, strong at polished documents, slides and spreadsheets.
Sep 25, 2026	Build plugins for Claude	Developers can publish plugins to the directory.
Sep 22, 2026	Claude Opus 5.5	Near the top model's quality at about 40% lower running cost than Opus 5; higher usage limits on paid plans.
Sep 16, 2026	Cowork and chat become one Claude; Docs, Slides and Design in every conversation	See section I.
Sep 15, 2026	Salesforce in Claude (beta)	Pipeline and accounts inside Claude for organizations Salesforce approves.
Sep 10, 2026	Smart reports (beta, Enterprise)	How your team actually uses Claude, and which repeated patterns deserve a shared skill.
Sep 1, 2026	Claude Fable 5.1 (and Mythos 5.1 for vetted programs)	The top model for knowledge work; the one I use for this guide.
Aug 25, 2026	Memory in Cowork, editable topics, sensitive-topics opt-in	One memory across chat and cloud tasks; you can see and edit what it holds.
Aug 6, 2026	Skill and plugin security scanning (beta, Enterprise)	Admins can scan third-party skills for malicious content.

Desktop app notes in the same window: Customize was redesigned (Sep 17) with separate Skills, Plugins and Connectors pages; Google Docs, Sheets and Slides creation arrived (Sep 29); per-computer folder permission modes (Oct 22). Standalone Claude Design at claude.ai/design closes on December 14, 2026; it now lives inside conversations.

V. Learn From the Source: Anthropic Academy

Free, official, and the first place to send a new team member. (Anthropic's learning site appears under two names at the moment, Anthropic Academy and Claude Academy; same courses.)

Course	Who it is for	What you get
Introduction to Claude Cowork	Anyone starting out	Thirteen short lessons plus a quiz: first task, projects and instructions, skills and plugins, Claude in Chrome and Microsoft 365, working safely, sharing with a team.
Claude 101	Every professional user	Everyday tasks, getting better results, projects, artifacts, skills, connecting tools. Certificate.
AI Fluency: Framework and Foundations	Everyone, about 4 hours	The 4D method: Delegation, Description, Discernment, Diligence. The best single course on thinking clearly about AI work.
AI Fluency for Small Businesses	Owners and staff	Applying the above to a small company.
Building Effective Human-Agent Teams (beta)	Leaders, 45 minutes	Moving a team from solo AI use to shared AI work.
Deploying Claude Enterprise with Confidence	Whoever leads a rollout	The five decisions: structure, access, governance, spend, visibility. Includes a work-along rollout plan.
Driving Enterprise Adoption of Claude	Project and enablement leads	Three phases (activation, acceleration, expansion), champions, train-the-trainer, and why an early productivity dip is normal.

For the technically curious: Introduction to Agent Skills, Introduction to Subagents, and Claude Code 101 sit on the same site.

VI. Tutorials Worth Your Time

Claude read the transcripts and checked the dates against Anthropic's own pages, and I review the list before each edition goes up. Claude changes fast, so anything recorded before September 16, 2026 shows screens that no longer exist; use those for ideas, not clicks.

No.	Watch	Who made it	When	Why it is on the list
I	Claude for Small Businesses (44 skills)	Ben AI	Sep 17, 2026, 14 min	Installs the small-business plugin, runs the onboarding step, then walks ten useful skills (inbox, reports, month-end, proposals, invoice chase, hiring). The closest thing to an owner's workflow tour. Skip the promotional segment.
II	8 Rules for Building Great Claude Skills	Ben AI	Oct 6, 2026, 15 min	Do the task first, keep skills small, add examples, load context only when needed, offer options at checkpoints. Practical and plain.
III	Claude Docs, Slides and Design, Full Guide	9x	Sep 18, 2026, 14 min	Click-by-click on the new outputs, sharing and comments, and the design-system step that stops everything looking generic. Recorded two days after the merge, so one menu path is already stale.
IV	5 Claude Cowork Tricks I Wish I Knew Day 1	Eliot Prince	Sep 28, 2026, 21 min	Guided setup, checking your usage, turning a long chat into a skill, phone alerts so long jobs do not stall. Honest about cost.
V	Claude Changed Completely: Every New Concept	Simon Scrapes	Oct 17, 2026, 19 min	The best recap of how thinking has shifted: skills as lists of corrections, defaults over tinkering, context in folders. Half of it is developer territory; the first half is for you.
VI	Claude Cowork Has a Massive Change	Ryan and Matt Data Science	Oct 5, 2026, 10 min	Explains the cloud move accurately, plus a reusable method for packaging a task's know-how into files so it survives any move.
VII	Claude Cowork and Chat Are Now One Claude	Claude (official)	Sep 16, 2026, 1 min	The official word on the merge. Not a tutorial, but authoritative.
VIII	Building Secure Agents for Knowledge Work	Claude (official) with Every	Oct 6, 2026, 34 min	For leaders: one shared company assistant, an editor's 30,000 past edits turned into a skill, judging AI on real work. Light on settings.
IX	Cowork Tutorial From Cowork's Design Lead	Peter Yang with Jenny Wen	Mar 29, 2026, 40 min	The intent behind the product: feed it messy sources, run tasks in parallel, schedule the weekly loop. Screens are out of date.

Skipped on purpose: the "full courses" of two to three hours (almost all pre-merge), anything titled "killed", "destroyed", "get rich" or "24/7 employee", and the many beginner walkthroughs recorded before September. Governance for non-technical teams is the thinnest area on YouTube; section VII fills that gap.

VII. Working Safely: The Short Checklist

Anthropic's own safety guidance, condensed for a firm that handles client data.

Give Claude one working folder, not your whole drive. Keep financial records, credentials and personal data out of it. Keep backups.
Use Manual approval for sensitive files, accounts, new tools and anything hard to reverse. Avoid "skip all approvals" for business work.
Every connector and plugin widens Claude's reach. Install only from sources you trust, read what a skill does before you turn it on, and remove what you do not use.
Watch for hijacking. Hidden instructions in a web page, an email or a document can steer Claude. The risk is highest when Claude can both read untrusted content and take actions. If it goes off topic, touches something unexpected or asks for sensitive information, stop and tell someone.
Keep the browser and computer use away from banking, healthcare and HR systems.
Start scheduled tasks on low-risk work and read the outputs for a few weeks before you widen them. Never schedule anything that sends messages or spends money without a review step.
Your firm is responsible for what Claude does on its behalf. Messages sent, purchases made, data changed. Treat it like a new employee with a company card.

For Team and Enterprise owners: cloud sessions, browser use, auto-approve, and write-capable connectors each have their own switch in organization settings, the most restrictive setting wins, plugins can be required, available or blocked by group, and the Compliance API and OpenTelemetry give you a record of what ran. Memory is off until you turn it on, and turning it off again deletes everyone's memory data.

VIII. About This Guide

I keep this page current so you do not have to. Every two weeks Claude re-reads Anthropic's release notes, changelog, and Academy, and re-vets the tutorials. I review each edition before it goes up here. What changed since the last edition appears at the top of section I.

Questions, corrections, or a tutorial I should look at: rlogan@placetostand.com.

Sources for this edition (all checked October 25, 2026)

Want Claude working in your firm?

I set it up around one workflow at a time. Anything that sends, posts, pays, files, or speaks for your firm waits for a person's yes.

See the 30-day deployment, or start smaller with the Diagnostic.

Book a 15-minute fit call →